AI Factory

Build a defective product sorting system at a low cost that can easily separate normal and defective products for any product. By applying the latest AI technology to many workplaces that employ people to sort out defective products, it can be improved with an unmanned system. There is no need to use expensive vision equipment for this. One Deep Runner is enough. Once you have photos of normal and defective products, you can easily learn a deep learning model and apply it to new products.

Applications

• Defect detection of factory products
• Product serial number recognition
• Fruit sorting system
• Fish sorting system

Configuration

Input/output

• Remote control functions

            Defect Rate Statistics Monitor

            Camera video monitor

            Training data capture command

• Real-time input/output (PLC)

            Real-time good/defective product inspection results

            Camera capture command

Features

• As the machine automatically learns and recognizes images of defective products, it can be optimized and applied to various fields.

• By recognizing HDMI-based high-quality camera images, you can detect subtle differences in products.

• The classification algorithm with the highest recognition rate is used. The recognition accuracy is not degraded in the computation process because it performs precise floating-point computations instead of integer computations that often result in arithmetic errors.

• In case of recognition in native resolution, the delay time between the image frame and the recognition result is only 0.03 seconds, so there is no production line delay.

• It provides a high-resolution recognition function that divides a single high-resolution product image into pieces, recognizes each and then integrates them. By recognizing every nook and cranny of the product, various defect patterns can be recognized at the same time.

• It is a compact device and can be mounted on a standard DIN rail, so it can be installed and used on the job site in the factory.

• By providing electrically isolated GPIO ports, it can be easily linked with a PLC in the factory that sends and receives 24V or 48V level signals. This allows PLC and AI devices to organically control the process in the factory.

• Because it embeds a web protocol, it can easily link with a remote server to report recognition results and manage settings remotely.

• Provides the function to automatically collect camera images as training data according to the instructions of the PLC or remote server.

• It provides a graphical interface-based deep learning training tool and can easily train a model in a short time using the new training data to improve recognition performance.

• You can build the system at a low cost.